Search NASA⌕ Search

SEARCH · Search NASA

Results for “Drug discovery”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Towards interpretable Cryo-EM: disentangling latent spaces of molecular conformations

Molecules are essential building blocks of life and their different conformations (i.e., shapes) crucially determine the functional role that they play in living organisms. Cryogenic Electron Microscopy (cryo-EM) allows for acquisition of large image datasets of individual molecules. Recent advances in computational cryo-EM have made it possible to learn latent variable models of conformation landscapes. However, interpreting these latent spaces remains a challenge as their individual dimensions are often arbitrary. The key message of our work is that this interpretation challenge can be viewed as an Independent Component Analysis (ICA) problem where we seek models that have the property of identifiability. That means, they have an essentially unique solution, representing a conformational latent space that separates the different degrees of freedom a molecule is equipped with in nature. Thus, we aim to advance the computational field of cryo-EM beyond visualizations as we connect it with the theoretical framework of (nonlinear) ICA and discuss the need for identifiable models, improved metrics, and benchmarks. Moving forward, we propose future directions for enhancing the disentanglement of latent spaces in cryo-EM, refining evaluation metrics and exploring techniques that leverage physics-based decoders of biomolecular systems. Moreover, we discuss how future technological developments in time-resolved single particle imaging may enable the application of nonlinear ICA models that can discover the true conformation changes of molecules in nature. The pursuit of interpretable conformational latent spaces will empower researchers to unravel complex biological processes and facilitate targeted interventions. This has significant implications for drug discovery and structural biology more broadly. More generally, latent variable models are deployed widely across many scientific disciplines. Thus, the argument we present in this work has much broader applications in AI for science if we want to move from impressive nonlinear neural network models to mathematically grounded methods that can help us learn something new about nature.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Protein-ligand binding affinity prediction using multi-instance learning with docking structures

Recent advances in 3D structure-based deep learning approaches demonstrate improved accuracy in predicting protein-ligand binding affinity in drug discovery. These methods complement physics-based computational modeling such as molecular docking for virtual high-throughput screening. Despite recent advances and improved predictive performance, most methods in this category primarily rely on utilizing co-crystal complex structures and experimentally measured binding affinities as both input and output data for model training. Nevertheless, co-crystal complex structures are not readily available and the inaccurate predicted structures from molecular docking can degrade the accuracy of the machine learning methods. We introduce a novel structure-based inference method utilizing multiple molecular docking poses for each complex entity. Our proposed method employs multi-instance learning with an attention network to predict binding affinity from a collection of docking poses. We validate our method using multiple datasets, including PDBbind and compounds targeting the main protease of SARS-CoV-2. The results demonstrate that our method leveraging docking poses is competitive with other state-of-the-art inference models that depend on co-crystal structures. This method offers binding affinity prediction without requiring co-crystal structures, thereby increasing its applicability to protein targets lacking such data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Preventing Failures By Dataset Shift Detection in Safety-Critical Graph Applications

Dataset shift refers to the problem where the input data distribution may change over time (e.g., between training and test stages). Since this can be a critical bottleneck in several safety-critical applications such as healthcare, drug-discovery, etc., dataset shift detection has become an important research issue in machine learning. Though several existing efforts have focused on image/video data, applications with graph-structured data have not received sufficient attention. Therefore, in this paper, we investigate the problem of detecting shifts in graph structured data through the lens of statistical hypothesis testing. Specifically, we propose a practical two-sample test based approach for shift detection in large-scale graph structured data. Our approach is very flexible in that it is suitable for both undirected and directed graphs, and eliminates the need for equal sample sizes. Using empirical studies, we demonstrate the effectiveness of the proposed test in detecting dataset shifts. We also corroborate these findings using real-world datasets, characterized by directed graphs and a large number of nodes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Improving Protein–Ligand Interaction Modeling with cryo-EM Data, Templates, and Deep Learning in 2021 Ligand Model Challenge

Elucidating protein–ligand interaction is crucial for studying the function of proteins and compounds in an organism and critical for drug discovery and design. The problem of protein–ligand interaction is traditionally tackled by molecular docking and simulation, which is based on physical forces and statistical potentials and cannot effectively leverage cryo-EM data and existing protein structural information in the protein–ligand modeling process. In this work, we developed a deep learning bioinformatics pipeline (DeepProLigand) to predict protein–ligand interactions from cryo-EM density maps of proteins and ligands. DeepProLigand first uses a deep learning method to predict the structure of proteins from cryo-EM maps, which is averaged with a reference (template) structure of the proteins to produce a combined structure to add ligands. The ligands are then identified and added into the structure to generate a protein–ligand complex structure, which is further refined. The method based on the deep learning prediction and template-based modeling was blindly tested in the 2021 EMDataResource Ligand Challenge and was ranked first in fitting ligands to cryo-EM density maps. These results demonstrate that the deep learning bioinformatics approach is a promising direction for modeling protein–ligand interactions on cryo-EM data using prior structural information.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

A Comprehensive Investigation of Active Learning Strategies for Conducting Anti-Cancer Drug Screening

It is well-known that cancers of the same histology type can respond differently to a treatment. Thus, computational drug response prediction is of paramount importance for both preclinical drug screening studies and clinical treatment design. To build drug response prediction models, treatment response data need to be generated through screening experiments and used as input to train the prediction models. In this study, we investigate various active learning strategies of selecting experiments to generate response data for the purposes of (1) improving the performance of drug response prediction models built on the data and (2) identifying effective treatments. Here, we focus on constructing drug-specific response prediction models for cancer cell lines. Various approaches have been designed and applied to select cell lines for screening, including a random, greedy, uncertainty, diversity, combination of greedy and uncertainty, sampling-based hybrid, and iteration-based hybrid approach. All of these approaches are evaluated and compared using two criteria: (1) the number of identified hits that are selected experiments validated to be responsive, and (2) the performance of the response prediction model trained on the data of selected experiments. The analysis was conducted for 57 drugs and the results show a significant improvement on identifying hits using active learning approaches compared with the random and greedy sampling method. Active learning approaches also show an improvement on response prediction performance for some of the drugs and analysis runs compared with the greedy sampling method.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Analysis of the Innovation Trend in Cell-Free Synthetic Biology

Cell-free synthetic biology is a maturing field that aims to assemble biomolecular reactions outside cells for compelling applications in drug discovery, metabolic engineering, biomanufacturing, diagnostics, and education. Cell-free systems have several key features. They circumvent mechanisms that have evolved to facilitate species survival, bypass limitations on molecular transport across the cell wall, enable high-yielding and rapid synthesis of proteins without creating recombinant cells, and provide high tolerance towards toxic substrates or products. Here, we analyze ~750 published patents and ~2000 peer-reviewed manuscripts in the field of cell-free systems. Three hallmarks emerged. First, we found that both patent filings and manuscript publications per year are significantly increasing (five-fold and 1.5-fold over the last decade, respectively). Second, we observed that the innovation landscape has changed. Patent applications were dominated by Japan in the early 2000s before shifting to China and the USA in recent years. Finally, we discovered an increasing prevalence of biotechnology companies using cell-free systems. Our analysis has broad implications on the future development of cell-free synthetic biology for commercial and industrial applications.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Nerpa: A Tool for Discovering Biosynthetic Gene Clusters of Bacterial Nonribosomal Peptides

Microbial natural products are a major source of bioactive compounds for drug discovery. Among these molecules, nonribosomal peptides (NRPs) represent a diverse class of natural products that include antibiotics, immunosuppressants, and anticancer agents. Recent breakthroughs in natural product discovery have revealed the chemical structure of several thousand NRPs. However, biosynthetic gene clusters (BGCs) encoding them are known only for a few hundred compounds. Here, we developed Nerpa, a computational method for the high-throughput discovery of novel BGCs responsible for producing known NRPs. After searching 13,399 representative bacterial genomes from the RefSeq repository against 8368 known NRPs, Nerpa linked 117 BGCs to their products. We further experimentally validated the predicted BGC of ngercheumicin from Photobacterium galatheae via mass spectrometry. Nerpa supports searching new genomes against thousands of known NRP structures, and novel molecular structures against tens of thousands of bacterial genomes. The availability of these tools can enhance our understanding of NRP synthesis and the function of their biosynthetic enzymes.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Challenges for Implementing FAIR Digital Objects with High Performance Workflows

New types of workflows are being used in science that couple traditional distributed and high-performance computing (HPC) with data-intensive approaches, and orchestrate ensembles of numerical simulations and artificial intelligence (AI) models. Such workflows may use AI models to supplement computation where numerical simulations may be too computationally expensive, to automate trivial yet time consuming operations, to perform preliminary selections among intractable numbers of combinations in domains as diverse as protein binding, fine-grid climate simulations, and drug discovery.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Inhibition of the proton-activated chloride channel PAC by PIP 2

Proton-activated chloride (PAC) channel is a ubiquitously expressed pH-sensing ion channel, encoded by PACC1 (TMEM206). PAC regulates endosomal acidification and macropinosome shrinkage by releasing chloride from the organelle lumens. It is also found at the cell surface, where it is activated under pathological conditions related to acidosis and contributes to acid-induced cell death. However, the pharmacology of the PAC channel is poorly understood. Here, we report that phosphatidylinositol (4,5)-bisphosphate (PIP 2 ) potently inhibits PAC channel activity. We solved the cryo-electron microscopy structure of PAC with PIP 2 at pH 4.0 and identified its putative binding site, which, surprisingly, locates on the extracellular side of the transmembrane domain (TMD). While the overall conformation resembles the previously resolved PAC structure in the desensitized state, the TMD undergoes remodeling upon PIP 2 -binding. Structural and electrophysiological analyses suggest that PIP 2 inhibits the PAC channel by stabilizing the channel in a desensitized-like conformation. Our findings identify PIP 2 as a new pharmacological tool for the PAC channel and lay the foundation for future drug discovery targeting this channel.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Identification of potent inhibitors of JUN N-terminal kinases for treatment of endometriosis and associated pain

Endometriosis, defined as the ectopic growth of endometrial tissue outside of the uterine cavity, is an inflammatory and hormone-dependent disease that causes excruciating pelvic pain, infertility, and significantly decreases quality of life in affected patients. The JUN N-terminal kinases (JNKs) are a leading class of nonhormonal therapeutic targets that have been validated in preclinical models of endometriosis and in a Phase 1/2 clinical trial. Despite their therapeutic potential, JNK inhibitors with increased potency and specificity are needed to address the inflammatory pathology of endometriosis and to prevent disease progression. Leveraging a DNA-encoded chemical library collection of ~4 billion compounds, we identified lead inhibitor CDD-2428 and optimized derivatives, CDD-2728 and CDD-3013, with excellent binding affinity to JNK1-3 (K d = 0.12 to 3.7 nM), enhanced selectivity, metabolic stability, and cellular permeability. Crystallographic and biochemical studies confirmed that CDD-3013 exhibited superior kinase selectivity with improved efficacy compared to existing JNK inhibitors. In primary endometriosis cell models, CDD-2728 and CDD-3013 suppressed JNK-dependent inflammatory signaling, dampening pathways linked to pain, invasion, angiogenesis, and macrophage recruitment. In an endometriosis mouse model, both CDD-2728 and CDD-3013 reduced endometriotic lesion size, macrophage infiltration, and cellular proliferation, showing in vivo efficacy. When tested in a lipopolysaccharide-induced hyperalgesia model, CDD-2728 and CDD-3013 decreased markers of induced pain, as measured by changes in a dynamic weight bearing test and Grimace scores. These findings nominate CDD-2728 and CDD-3013 as potent, nonhormonal therapeutic candidates for endometriosis with broad anti-inflammatory and analgesic activity, addressing a critical unmet clinical need.

Madasu, Chandrashekhar [Department of Pathology an↗
DOE OSTI2026

Exploration of Nirmatrelvir Derivatives as Optimized SARS‐CoV‐2 Antivirals

Nirmatrelvir (NMV) is a SARS‐CoV‐2 antiviral component of the approved COVID‐19 therapeutic Paxlovid. It is a reversible covalent inhibitor of SARS‐CoV‐2 main protease (M Pro ) that is effluxed from human cells by P‐glycoprotein (P‐gp). To identify NMV analogs with improved potency and reduced P‐gp efflux, a structure–activity relationship campaign was conducted. Warheads alternative to nitrile for engaging the active site cysteine were tested showing aldehyde and dichloroacetamide with better enzyme inhibition potency. Crystal structure of MPI‐136−M Pro shows its aldehyde warhead forming a thiohemiacetal with active Cys145 of M Pro . Several S4 binders were explored revealing that an O‐to‐S shift at the N ‐terminal amide leads to better enzyme inhibition. By exploring different combinations of S2, S3, and S4 binders, two inhibitors with better enzyme inhibition potency than NMV were found. Crystal structure of MPI‐148, with ( S )‐2‐azaspiro[4,5]decane‐3‐carboxylate as an alternative S2 binder, shows extensive hydrogen‐bond networks for locking the inhibitor in active site, explaining high affinity of NMV analogs. Further characterization of cellular M Pro engagement and antiviral potency against SARS‐CoV‐2 revealed four inhibitors with greater potency than NMV in P‐gp‐expressing cells. Studies with the P‐gp inhibitor CP‐100356 showed that these compounds were less sensitive to P‐gp inhibition than NMV, consistent with reduced P‐gp‐mediated efflux.

Alugubelli, Yugendar R. [Texas A&M Drug Discovery ↗
DOE OSTI2021

Artificial Intelligence for Autonomous Molecular Design: A Perspective

Domain-aware artificial intelligence has been increasingly adopted in recent years to expedite molecular design in various applications, including drug design and discovery. Recent advances in areas such as physics-informed machine learning and reasoning, software engineering, high-end hardware development, and computing infrastructures are providing opportunities to build scalable and explainable AI molecular discovery systems. This could improve a design hypothesis through feedback analysis, data integration that can provide a basis for the introduction of end-to-end automation for compound discovery and optimization, and enable more intelligent searches of chemical space. Several state-of-the-art ML architectures are predominantly and independently used for predicting the properties of small molecules, their high throughput synthesis, and screening, iteratively identifying and optimizing lead therapeutic candidates. However, such deep learning and ML approaches also raise considerable conceptual, technical, scalability, and end-to-end error quantification challenges, as well as skepticism about the current AI hype to build automated tools. To this end, synergistically and intelligently using these individual components along with robust quantum physics-based molecular representation and data generation tools in a closed-loop holds enormous promise for accelerated therapeutic design to critically analyze the opportunities and challenges for their more widespread application. This article aims to identify the most recent technology and breakthrough achieved by each of the components and discusses how such autonomous AI and ML workflows can be integrated to radically accelerate the protein target or disease model-based probe design that can be iteratively validated experimentally. Taken together, this could significantly reduce the timeline for end-to-end therapeutic discovery and optimization upon the arrival of any novel zoonotic transmission event. Our article serves as a guide for medicinal, computational chemistry and biology, analytical chemistry, and the ML community to practice autonomous molecular design in precision medicine and drug discovery.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Novel small molecule FGF 23 inhibitors increase serum phosphate and improve skeletal abnormalities in Hyp mice

We report that excess fibroblast growth factor 23 (FGF23) causes hereditary hypophosphatemic rickets, such as X-linked hypophosphatemia (XLH) and tumor induced osteomalacia (TIO). A small molecule that specifically binds to FGF23 to prevent activation of the FGFR/a-Klotho complex has potential advantages over the currently approved systemically administered FGF23 blocking antibody. Using structure-based drug design we previously identified ZINC13407541 (N-[[2-(2-phenylethenyl)cyclopenten-1-yl]methylidene]hydroxylamine) as a small molecule antagonist for FGF23. Additional structure-activity studies developed a series of ZINC13407541 analogues with enhanced drug-like properties. In this study, we tested in a pre-clinical Hyp mouse homologue of XLH a direct connect analogue (8n) [(E)-2-(4-(tert-butyl)phenyl)cyclopent-1-ene-1-carbaldehyde oxime] that exhibited the greatest stability in microsomal assays, and 13a [(E)-2-((E)-4-methylstyryl)benzaldehyde oxime] that exhibited increased in vitro potency. Using cryo-electron microscopy (Cryo-EM) structure and computational docking, we identified a key binding residue (Q156) of the FGF23 antagonists, ZINC13407541 and its analogues (8n and 13a) in the N-terminal domain of FGF23 protein. Site-directed mutagenesis and bimolecular fluorescence complementation (BiFC)-fluorescence resonance energy transfer (FRET) assay confirmed the binding site of these three antagonists. We found that pharmacological inhibition of FGF23 with either of these compounds blocked FGF23 signaling and increased serum phosphate and 1,25(OH)2D concentrations in Hyp mice. Long-term parenteral treatment with 8n or 13a also enhanced linear bone growth, increased mineralization of bone, and narrowed the growth plate in Hyp mice. The more potent 13a compound had greater therapeutic effects in Hyp mice. Further optimization of these FGF23 inhibitors may lead to versatile drugs to treat excess FGF23-mediated disorders.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2023

Potent and selective covalent inhibition of the papain-like protease from SARS-CoV-2

Abstract Direct-acting antivirals are needed to combat coronavirus disease 2019 (COVID-19), which is caused by severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2). The papain-like protease (PLpro) domain of Nsp3 from SARS-CoV-2 is essential for viral replication. In addition, PLpro dysregulates the host immune response by cleaving ubiquitin and interferon-stimulated gene 15 protein from host proteins. As a result, PLpro is a promising target for inhibition by small-molecule therapeutics. Here we design a series of covalent inhibitors by introducing a peptidomimetic linker and reactive electrophile onto analogs of the noncovalent PLpro inhibitor GRL0617. The most potent compound inhibits PLpro with k inact /K I = 9,600 M −1 s −1 , achieves sub-μM EC 50 values against three SARS-CoV-2 variants in mammalian cell lines, and does not inhibit a panel of human deubiquitinases (DUBs) at >30 μM concentrations of inhibitor. An X-ray co-crystal structure of the compound bound to PLpro validates our design strategy and establishes the molecular basis for covalent inhibition and selectivity against structurally similar human DUBs. These findings present an opportunity for further development of covalent PLpro inhibitors.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Stereoselective synthesis of a 4-α-glucoside of valienamine and its X-ray structure in complex with Streptomyces coelicolor GlgE1-V279S

Glycoside hydrolases (GH) are a large family of hydrolytic enzymes found in all domains of life. As such, they control a plethora of normal and pathogenic biological functions. Thus, understanding selective inhibition of GH enzymes at the atomic level can lead to the identification of new classes of therapeutics. In these studies, we identified a 4-α-glucoside of valienamine (8) as an inhibitor of Streptomyces coelicolor (Sco) GlgE1-V279S which belongs to the GH13 Carbohydrate Active EnZyme family. The results obtained from the dose–response experiments show that 8 at a concentration of 1000 µM reduced the enzyme activity of Sco GlgE1-V279S by 65%. The synthetic route to 8 and a closely related 4-α-glucoside of validamine (7) was achieved starting from readily available D-maltose. A key step in the synthesis was a chelation-controlled addition of vinylmagnesium bromide to a maltose-derived enone intermediate. X-ray structures of both 7 and 8 in complex with Sco GlgE1-V279S were solved to resolutions of 1.75 and 1.83 Å, respectively. Structural analysis revealed the valienamine derivative 8 binds the enzyme in an E 2 conformation for the cyclohexene fragment. Also, the cyclohexene fragment shows a new hydrogen-bonding contact from the pseudo-diaxial C(3)–OH to the catalytic nucleophile Asp 394 at the enzyme active site. Asp 394, in fact, forms a bidentate interaction with both the C(3)–OH and C(7)-OH of the inhibitor. In contrast, compound 7 disrupts the catalytic sidechain interaction network of Sco GlgE1-V279S via steric interactions resulting in a conformation change in Asp 394. These findings will have implications for the design other aminocarbasugar-based GH13-inhibitors and will be useful for identifying more potent and selective inhibitors.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Structural basis for SHOC2 modulation of RAS signalling

The RAS–RAF pathway is one of the most commonly dysregulated in human cancers. Despite decades of study, understanding of the molecular mechanisms underlying dimerization and activation of the kinase RAF remains limited. Recent structures of inactive RAF monomer and active RAF dimer bound to 14-3-3 have revealed the mechanisms by which 14-3-3 stabilizes both RAF conformations via specific phosphoserine residues. Prior to RAF dimerization, the protein phosphatase 1 catalytic subunit (PP1C) must dephosphorylate the N-terminal phosphoserine (NTpS) of RAF to relieve inhibition by 14-3-3, although PP1C in isolation lacks intrinsic substrate selectivity. SHOC2 is as an essential scaffolding protein that engages both PP1C and RAS to dephosphorylate RAF NTpS, but the structure of SHOC2 and the architecture of the presumptive SHOC2–PP1C–RAS complex remain unknown. Here we present a cryo-electron microscopy structure of the SHOC2–PP1C–MRAS complex to an overall resolution of 3Å, revealing a tripartite molecular architecture in which a crescent-shaped SHOC2 acts as a cradle and brings together PP1C and MRAS. Our work demonstrates the GTP dependence of multiple RAS isoforms for complex formation, delineates the RAS-isoform preference for complex assembly, and uncovers how the SHOC2 scaffold and RAS collectively drive specificity of PP1C for RAF NTpS. Our data indicate that disease-relevant mutations affect complex assembly, reveal the simultaneous requirement of two RAS molecules for RAF activation, and establish rational avenues for discovery of new classes of inhibitors to target this pathway.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Iterative computational design and crystallographic screening identifies potent inhibitors targeting the Nsp3 macrodomain of SARS-CoV-2

The nonstructural protein 3 (NSP3) of the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) contains a conserved macrodomain enzyme (Mac1) that is critical for pathogenesis and lethality. While small-molecule inhibitors of Mac1 have great therapeutic potential, at the outset of the COVID-19 pandemic, there were no well-validated inhibitors for this protein nor, indeed, the macrodomain enzyme family, making this target a pharmacological orphan. Here, we report the structure-based discovery and development of several different chemical scaffolds exhibiting low- to sub-micromolar affinity for Mac1 through iterations of computer-aided design, structural characterization by ultra-high-resolution protein crystallography, and binding evaluation. Potent scaffolds were designed with in silico fragment linkage and by ultra-large library docking of over 450 million molecules. Both techniques leverage the computational exploration of tangible chemical space and are applicable to other pharmacological orphans. Overall, 160 ligands in 119 different scaffolds were discovered, and 153 Mac1-ligand complex crystal structures were determined, typically to 1 Å resolution or better. Our analyses discovered selective and cell-permeable molecules, unexpected ligand-mediated conformational changes within the active site, and key inhibitor motifs that will template future drug development against Mac1.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

CACTUS: Chemistry Agent Connecting Tool Usage to Science

Large language models (LLMs) have shown remarkable potential in various domains but often lack the ability to access and reason over domain-specific knowledge and tools. In this article, we introduce Chemistry Agent Connecting Tool-Usage to Science (CACTUS), an LLM-based agent that integrates existing cheminformatics tools to enable accurate and advanced reasoning and problem-solving in chemistry and molecular discovery. We evaluate the performance of CACTUS using a diverse set of open-source LLMs, including Gemma-7b, Falcon-7b, MPT-7b, Llama3-8b, and Mistral-7b, on a benchmark of thousands of chemistry questions. Our results demonstrate that CACTUS significantly outperforms baseline LLMs, with the Gemma-7b, Mistral-7b, and Llama3-8b models achieving the highest accuracy regardless of the prompting strategy used. Moreover, we explore the impact of domain-specific prompting and hardware configurations on model performance, highlighting the importance of prompt engineering and the potential for deploying smaller models on consumer-grade hardware without a significant loss in accuracy. By combining the cognitive capabilities of open-source LLMs with widely used domain-specific tools provided by RDKit, CACTUS can assist researchers in tasks such as molecular property prediction, similarity searching, and drug-likeness assessment.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗